A5001004876- Natural Language Processing Techniques
- Topic Modeling
- Text Readability and Simplification
- Diabetes Management and Education
- Linguistics, Language Diversity, and Identity
- Diabetes Management and Research
- Lexicography and Language Studies
- Information Retrieval and Search Behavior
- Diabetes and associated disorders
- Network Security and Intrusion Detection
- Network Traffic and Congestion Control
- Multimodal Machine Learning Applications
- Machine Learning and Algorithms
- Internet Traffic Analysis and Secure E-voting
- Recommender Systems and Techniques
- Speech Recognition and Synthesis
- Text and Document Classification Technologies
Microsoft (India)
2018-2020
Birla Institute of Technology and Science, Pilani - Goa Campus
2015
Type 1 diabetes (T1D) is emerging as a major healthcare challenge impacting significant percentage of the population. Management T1D remains concern in India with diversity socio-economic backgrounds, poor literacy levels, and inadequate resources facilities which impact timely diagnosis, treatment, management this condition. All over world general, India, many peer support groups have come into existence are even blossoming. A system instils sense security, allows for better care practices,...
A network with dynamic weights implies a set of vertices interconnected by edges, each which bears weight that changes time. One example such networks is traffic network, wherein, the structure graph remains constant but on signifying amount (traffic density) over We have dealt scenarios where flow algorithm needs to run repeatedly establish flows in timely changing capacities and we sought obtain some form computational intelligence subject. aligned our work context order explore practical...
In contemporary machine learning approaches to bilingual lexicon induction (BLI), a model learns mapping between the embedding spaces of language pair. Recently, retrieve-and-rank approach BLI has achieved state art results on task. However, problem remains challenging in low-resource settings, due paucity data. The task is complicated by factors such as lexical variation across languages. We argue that incorporation additional information into recent should improve induction. demonstrate...
Document categorization is the process of assigning pre-defined categories to textual documents. State-of-the art approaches have modelled documents in terms corpus-length long vectors and viewed problem only from a syntactic perspective. We develop general measure estimate semantic closeness by utilizing relatedness most discriminative individual words that define document. Anaphora resolution used strengthen meaning ascribed each Our framework benefits word semantics Wordnet taxonomy thus...
Search queries issued over the Web increasingly look like questions, especially as domain becomes more specific. Finding good response to such amounts finding relevant passages from documents. Traditional information retrieval based search still matches query words in entire document. With advent of machine reading comprehension techniques, is moving towards identifying best sentence / group sentences We present AQuPR an A ttention Qu ery P assage R etrieval system find human acceptable...
Related or ideal follow-up suggestions to a web query in search engines are often optimized based on several different parameters -- relevance the original query, diversity, click probability etc. One many rankers may be trained score each suggestion from candidate pool these factors. These scorers usually pairwise classification tasks where training example consists of user and single list candidates. We propose an architecture that takes all associated with given outputs block. discuss...
Large Scale Question-Answering systems today are widely used in downstream applications such as chatbots and conversational dialogue agents. Typically, consist of an Answer Passage retrieval layer coupled with Machine Comprehension models trained on natural language query-passage pairs. Recent studies have explored Question Answering over structured data sources web-tables relational databases. However, architectures Seq2SQL assume the correct table a priori which is input to model along...
Modern transformer-based neural architectures yield impressive results in nearly every NLP task and Word Sense Disambiguation, the problem of discerning correct sense a word given context, is no exception. State-of-the-art approaches WSD today leverage lexical information along with pre-trained embeddings from these models to achieve comparable human inter-annotator agreement on standard evaluation benchmarks. In same vein, we experiment several strategies optimize bi-encoders for this...